Table 3. Comparison of average metrics among various architectures.

The bold represents the two best values of the metrics (highest accuracy, lowest loss) for the evaluated architectures.

CNN model	Training		Validation		Testing
	Accuracy(%)	Loss (×10−3)	Accuracy(%)	Loss (×10−3)	Accuracy(%)	Loss (×10−3)
32 filters. 1 FC 128 neurons	94.26	15.17	90.00	22.88	87.06	53.69
32 filters. 1 FC 64 neurons	94.65	15.87	90.00	33.65	84.71	44.56
32 filters. 2 FC 64 neurons	88.71	26.96	90.00	26.41	80.00	58.95
32 filters. 2 FC 128 neurons	94.26	15.62	94.00	24.72	92.94	26.38
64 filters. 1 FC 64 neurons	61.98	66.47	66.47	66.31	62.35	66.31
64 filters. 1 FC 128 neurons	94.06	15.57	94.00	20.36	90.59	25.86
64 filters. 2 FC 64 neurons	95.45	11.66	90.00	25.11	89.41	52.80
64 filters. 2FC 128 neurons	86.53	33.56	84.00	40.77	76.47	47.99